markstream-vue2-cli
Warn
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to perform installation of third-party packages using package managers (e.g.,
npm install markstream-vue2). - [EXTERNAL_DOWNLOADS]: The workflow suggests using CDN-hosted workers for rendering KaTeX and Mermaid diagrams when local worker imports are unavailable. This introduces a dependency on remote scripts served from external infrastructure.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and render Markdown content (e.g., chat history or streaming responses). This presents an injection surface where untrusted data could influence agent behavior or compromise the frontend environment if not properly sanitized.
- Ingestion points: Data passed to the
:contentprop in theMarkdownRendercomponent (documented in SKILL.md). - Boundary markers: No specific boundary markers or delimiters for untrusted content are implemented in the provided code example.
- Capability inventory: The skill enables rendering of Markdown, HTML, and Mermaid diagrams, which can include link execution and complex script-like diagram logic.
- Sanitization: The instructions include advisory notes to "Keep HTML safe and Mermaid strict" and validate against untrusted Markdown, though implementation details are left to the user.
- [DYNAMIC_EXECUTION]: The integration utilizes Web Workers (via CDN) to dynamically render diagrams and math formulas. This involves executing logic within worker threads based on the provided Markdown input string.
- [AUTONOMY_CONCERNS]: The skill includes instructions that a "requested scoped integration already authorizes routine source/dependency edits," which may encourage the agent to skip user confirmation for certain file modifications within the project scope.
Audit Metadata